The purposes of this study were to determine for signal-averaged electrocardiograms in normal subjects: (1) the incidence of false positives for various filter frequencies and late potential criteria and (2) reproducibility over time. In 46 normal volunteers, the QRS vector magnitude was bidirectionally high-pass filtered at 25, 40, and 100 Hz. As high-pass filter frequency increased, QRS duration decreased from 98 +/- 9 to 92 +/- 9 msec (p less than 0.0001), terminal QRS root mean square voltage decreased from 60 +/- 41 to 14 +/- 9 microV (p less than 0.0001), and terminal QRS low amplitude signal duration increased from 27 +/- 7 to 41 +/- 14 msec (p less than 0.0001). For individual parameters, the incidence of false positive tests ranged from 2% to 41%, whereas there were no false positive tests for the combination of abnormal QRS duration plus either root mean square voltage or low amplitude signal duration. Measurements were repeated after 6.4 +/- 0.3 months in 26 subjects and were highly reproducible at all filter settings. The potentially high incidence of false positive tests with some criteria has important implications for the use of signal-averaged electrocardiography as a screening test in patient populations with various arrhythmia risks.